add_rocm_support

Signed-off-by: Song <[email protected]>

README.md

@@ -71,7 +71,6 @@ Recently the state space models (SSMs) with efficient hardware-aware designs, i.
 | [Vim-tiny<sup>+</sup>](https://huggingface.co/hustvl/Vim-tiny-midclstok)    |       7M       |   78.3   | 94.2 | https://huggingface.co/hustvl/Vim-tiny-midclstok |
 | [Vim-small](https://huggingface.co/hustvl/Vim-small-midclstok)    |       26M       |   80.5   | 95.1 | https://huggingface.co/hustvl/Vim-small-midclstok |
 | [Vim-small<sup>+</sup>](https://huggingface.co/hustvl/Vim-small-midclstok)    |       26M       |   81.6   | 95.4 | https://huggingface.co/hustvl/Vim-small-midclstok |
-| [Vim-base](https://huggingface.co/hustvl/Vim-base-midclstok)    |       98M       |   81.9   | 95.8 | https://huggingface.co/hustvl/Vim-base-midclstok |
 
 **Notes:**
 - <sup>+</sup> means that we finetune at finer granularity with short schedule.

causal-conv1d/README.md

@@ -1 +1,43 @@
 # Causal depthwise conv1d in CUDA with a PyTorch interface
+
+Features:
+- Support fp32, fp16, bf16.
+- Kernel size 2, 3, 4.
+
+## How to use
+
+```
+from causal_conv1d import causal_conv1d_fn
+```
+
+```
+def causal_conv1d_fn(x, weight, bias=None, activation=None):
+    """
+    x: (batch, dim, seqlen)
+    weight: (dim, width)
+    bias: (dim,)
+    activation: either None or "silu" or "swish"
+
+    out: (batch, dim, seqlen)
+    """
+```
+
+Equivalent to:
+```
+import torch.nn.functional as F
+
+F.conv1d(x, weight.unsqueeze(1), bias, padding=width - 1, groups=dim)[..., :seqlen]
+```
+
+## Additional Prerequisites for AMD cards
+
+### Patching ROCm
+
+If you are on ROCm 6.0, run the following steps to avoid errors during compilation. This is not required for ROCm 6.1 onwards.
+
+1. Locate your ROCm installation directory. This is typically found at `/opt/rocm/`, but may vary depending on your installation.
+
+2. Apply the Patch. Run with `sudo` in case you encounter permission issues.
+   ```bash
+    patch /opt/rocm/include/hip/amd_detail/amd_hip_bf16.h < rocm_patch/rocm6_0.patch 
+   ```

causal-conv1d/causal_conv1d/__init__.py

@@ -1,3 +1,3 @@
-__version__ = "1.0.0"
+__version__ = "1.4.0"
 
 from causal_conv1d.causal_conv1d_interface import causal_conv1d_fn, causal_conv1d_update

causal-conv1d/causal_conv1d/causal_conv1d_interface.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2023, Tri Dao.
+# Copyright (c) 2024, Tri Dao.
 
 import torch
 import torch.nn.functional as F
@@ -9,48 +9,142 @@
 
 class CausalConv1dFn(torch.autograd.Function):
     @staticmethod
-    def forward(ctx, x, weight, bias=None, activation=None):
+    def forward(
+        ctx,
+        x,
+        weight,
+        bias=None,
+        seq_idx=None,
+        initial_states=None,
+        return_final_states=False,
+        final_states_out=None,
+        activation=None,
+    ):
         if activation not in [None, "silu", "swish"]:
             raise NotImplementedError("activation must be None, silu, or swish")
         if x.stride(2) != 1 and x.stride(1) != 1:
             x = x.contiguous()
         bias = bias.contiguous() if bias is not None else None
-        ctx.save_for_backward(x, weight, bias)
+        if seq_idx is not None:
+            assert (
+                initial_states is None
+            ), "initial_states must be None if seq_idx is not None"
+            assert (
+                not return_final_states
+            ), "If seq_idx is not None, we don't return final_states_out"
+        seq_idx = seq_idx.contiguous() if seq_idx is not None else None
+        if initial_states is not None and (
+            initial_states.stride(2) != 1 and initial_states.stride(1) != 1
+        ):
+            initial_states = initial_states.contiguous()
+        if return_final_states:
+            assert (
+                x.stride(1) == 1
+            ), "Only channel-last layout support returning final_states_out"
+            if final_states_out is not None:
+                assert (
+                    final_states_out.stride(2) == 1 or final_states_out.stride(1) == 1
+                )
+            else:
+                batch, dim, seqlen = x.shape
+                width = weight.shape[1]
+                final_states_out = torch.empty(
+                    batch, width - 1, dim, device=x.device, dtype=x.dtype
+                ).transpose(1, 2)
+        else:
+            final_states_out = None
         ctx.activation = activation in ["silu", "swish"]
-        out = causal_conv1d_cuda.causal_conv1d_fwd(x, weight, bias, ctx.activation)
-        return out
+        out = causal_conv1d_cuda.causal_conv1d_fwd(
+            x, weight, bias, seq_idx, initial_states, final_states_out, ctx.activation
+        )
+        ctx.save_for_backward(x, weight, bias, seq_idx, initial_states)
+        ctx.return_final_states = return_final_states
+        ctx.return_dinitial_states = (
+            initial_states is not None and initial_states.requires_grad
+        )
+        return out if not return_final_states else (out, final_states_out)
 
     @staticmethod
-    def backward(ctx, dout):
-        x, weight, bias = ctx.saved_tensors
+    def backward(ctx, dout, *args):
+        x, weight, bias, seq_idx, initial_states = ctx.saved_tensors
+        dfinal_states = args[0] if ctx.return_final_states else None
         if dout.stride(2) != 1 and dout.stride(1) != 1:
             dout = dout.contiguous()
         # The kernel supports passing in a pre-allocated dx (e.g., in case we want to fuse the
         # backward of conv1d with the backward of chunk).
         # Here we just pass in None and dx will be allocated in the C++ code.
-        dx, dweight, dbias = causal_conv1d_cuda.causal_conv1d_bwd(
-            x, weight, bias, dout, None, ctx.activation
+        dx, dweight, dbias, dinitial_states = causal_conv1d_cuda.causal_conv1d_bwd(
+            x,
+            weight,
+            bias,
+            dout,
+            seq_idx,
+            initial_states,
+            dfinal_states,
+            None,
+            ctx.return_dinitial_states,
+            ctx.activation,
+        )
+        return (
+            dx,
+            dweight,
+            dbias if bias is not None else None,
+            None,
+            dinitial_states if initial_states is not None else None,
+            None,
+            None,
+            None,
         )
-        return dx, dweight, dbias if bias is not None else None, None
 
 
-def causal_conv1d_fn(x, weight, bias=None, activation=None):
+def causal_conv1d_fn(
+    x,
+    weight,
+    bias=None,
+    seq_idx=None,
+    initial_states=None,
+    return_final_states=False,
+    final_states_out=None,
+    activation=None,
+):
     """
     x: (batch, dim, seqlen)
     weight: (dim, width)
     bias: (dim,)
+    seq_idx: (batch, seqlen)
+    initial_states: (batch, dim, width - 1)
+    final_states_out: (batch, dim, width - 1), to be written to
     activation: either None or "silu" or "swish"
 
     out: (batch, dim, seqlen)
     """
-    return CausalConv1dFn.apply(x, weight, bias, activation)
-
-
-def causal_conv1d_ref(x, weight, bias=None, activation=None):
+    return CausalConv1dFn.apply(
+        x,
+        weight,
+        bias,
+        seq_idx,
+        initial_states,
+        return_final_states,
+        final_states_out,
+        activation,
+    )
+
+
+def causal_conv1d_ref(
+    x,
+    weight,
+    bias=None,
+    initial_states=None,
+    return_final_states=False,
+    final_states_out=None,
+    activation=None,
+):
     """
     x: (batch, dim, seqlen)
     weight: (dim, width)
     bias: (dim,)
+    initial_states: (batch, dim, width - 1)
+    final_states_out: (batch, dim, width - 1)
 
     out: (batch, dim, seqlen)
     """
@@ -60,45 +154,86 @@ def causal_conv1d_ref(x, weight, bias=None, activation=None):
     x = x.to(weight.dtype)
     seqlen = x.shape[-1]
     dim, width = weight.shape
-    out = F.conv1d(x, weight.unsqueeze(1), bias, padding=width - 1, groups=dim)
+    if initial_states is None:
+        out = F.conv1d(x, weight.unsqueeze(1), bias, padding=width - 1, groups=dim)
+    else:
+        x = torch.cat([initial_states, x], dim=-1)
+        out = F.conv1d(x, weight.unsqueeze(1), bias, padding=0, groups=dim)
     out = out[..., :seqlen]
-    return (out if activation is None else F.silu(out)).to(dtype=dtype_in)
-
-
-def causal_conv1d_update(x, conv_state, weight, bias=None, activation=None):
+    if return_final_states:
+        final_states = F.pad(x, (width - 1 - x.shape[-1], 0)).to(
+            dtype_in
+        )  # (batch, dim, width - 1)
+        if final_states_out is not None:
+            final_states_out.copy_(final_states)
+        else:
+            final_states_out = final_states
+    out = (out if activation is None else F.silu(out)).to(dtype=dtype_in)
+    return out if not return_final_states else (out, final_states_out)
+
+
+def causal_conv1d_update(x, conv_state, weight, bias=None, activation=None, cache_seqlens=None):
     """
-    x: (batch, dim)
-    conv_state: (batch, dim, width)
+    x: (batch, dim) or (batch, dim, seqlen)
+    conv_state: (batch, dim, state_len), where state_len >= width - 1
     weight: (dim, width)
     bias: (dim,)
+    cache_seqlens: (batch,), dtype int32.
+        If not None, the conv_state is treated as a circular buffer.
+        The conv_state will be updated by copying x to the conv_state starting at the index
+        @cache_seqlens % state_len.
 
-    out: (batch, dim)
+    out: (batch, dim) or (batch, dim, seqlen)
     """
     if activation not in [None, "silu", "swish"]:
         raise NotImplementedError("activation must be None, silu, or swish")
     activation = activation in ["silu", "swish"]
-    return causal_conv1d_cuda.causal_conv1d_update(x, conv_state, weight, bias, activation)
-
-
-def causal_conv1d_update_ref(x, conv_state, weight, bias=None, activation=None):
+    unsqueeze = x.dim() == 2
+    if unsqueeze:
+        x = x.unsqueeze(-1)
+    out = causal_conv1d_cuda.causal_conv1d_update(
+        x, conv_state, weight, bias, activation, cache_seqlens
+    )
+    if unsqueeze:
+        out = out.squeeze(-1)
+    return out
+
+
+def causal_conv1d_update_ref(x, conv_state, weight, bias=None, activation=None, cache_seqlens=None):
     """
-    x: (batch, dim)
-    conv_state: (batch, dim, width)
+    x: (batch, dim) or (batch, dim, seqlen)
+    conv_state: (batch, dim, state_len), where state_len >= width - 1
     weight: (dim, width)
     bias: (dim,)
+    cache_seqlens: (batch,), dtype int32.
+        If not None, the conv_state is treated as a circular buffer.
+        The conv_state will be updated by copying x to the conv_state starting at the index
+        @cache_seqlens % state_len before performing the convolution.
 
-    out: (batch, dim)
+    out: (batch, dim) or (batch, dim, seqlen)
     """
     if activation not in [None, "silu", "swish"]:
         raise NotImplementedError("activation must be None, silu, or swish")
     dtype_in = x.dtype
-    batch, dim = x.shape
+    unsqueeze = x.dim() == 2
+    if unsqueeze:
+        x = x.unsqueeze(-1)
+    batch, dim, seqlen = x.shape
     width = weight.shape[1]
-    assert conv_state.shape == (batch, dim, width)
+    state_len = conv_state.shape[-1]
+    assert conv_state.shape == (batch, dim, state_len)
     assert weight.shape == (dim, width)
-    conv_state.copy_(torch.roll(conv_state, shifts=-1, dims=-1)) # Update state (B D W)
-    conv_state[:, :, -1] = x
-    out = torch.sum(conv_state * weight, dim=-1) # (B D)
-    if bias is not None:
-        out += bias
+    if cache_seqlens is None:
+        x_new = torch.cat([conv_state, x], dim=-1).to(weight.dtype)  # (batch, dim, state_len + seqlen)
+        conv_state.copy_(x_new[:, :, -state_len:])
+    else:
+        width_idx = torch.arange(-(width - 1), 0, dtype=torch.long, device=x.device).unsqueeze(0) + cache_seqlens.unsqueeze(1)
+        width_idx = torch.remainder(width_idx, state_len).unsqueeze(1).expand(-1, dim, -1)
+        x_new = torch.cat([conv_state.gather(2, width_idx), x], dim=-1).to(weight.dtype)
+        copy_idx = torch.arange(seqlen, dtype=torch.long, device=x.device).unsqueeze(0) + cache_seqlens.unsqueeze(1)
+        copy_idx = torch.remainder(copy_idx, state_len).unsqueeze(1).expand(-1, dim, -1)
+        conv_state.scatter_(2, copy_idx, x)
+    out = F.conv1d(x_new, weight.unsqueeze(1), bias, padding=0, groups=dim)[:, :, -seqlen:]
+    if unsqueeze:
+        out = out.squeeze(-1)
     return (out if activation is None else F.silu(out)).to(dtype=dtype_in)